Understanding and Enhancing Queries

万维网搜索者经常因为缺乏相关的结果而感到沮丧，或者更经常的是，被太大而无法检查相关性的结果集所淹没。每天执行数亿次查询，查询日志为该项目提供了新的知识来源，以了解用户如何搜索，人们正在搜索什么，并为未来的搜索者提供建议。该项目的目标是了解人们在Web上搜索什么;为不确定或缺乏经验的搜索者提供查询建议;并为网站的搜索引擎优化提供相关查询条件。为此，分析查询及其结果的二分图以识别有用的查询-查询和文档-文档关系;使用新的关系以及更传统的信息源将查询聚类成主题;现有的信息检索技术适于帮助识别、组织和跟踪不仅仅是查询流行度，而是主题流行度;并且利用来自查询日志的信息来帮助找到表达类似信息需求的优选查询。这项工作的广泛影响是双重的。通过利用Web搜索引擎查询跟踪中存在的未开发信息，该项目增加了人们如何在Web上搜索以及他们正在寻找什么的理解。这些知识被应用于生成算法和工具，以支持搜索者以及那些希望被搜索者找到的人。这些工具以及收集或生成的数据集将通过项目网站（http：//www.cse.lehigh.edu/tubrian/nsf/queries-03.html）提供给研究界。

World-Wide Web searchers are often frustrated by a lack of relevant results, or more often, overwhelmed by a result set that is too large to examine for relevancy. With hundreds of millions of queries performed each day, query logs provide a new source of knowledge for this project to learn how users search, what people are searching for, and provide suggestions to future searchers. The goals of this project are to learn what people search for on the Web; to provide query suggestions for uncertain or inexperienced searchers; and, to offer relevant query terms for search engine optimization of a website. To these ends, the bipartite graph of queries and their results are analyzed to identify useful query-query and document-document relationships; queries are clustered into topics, using the new relationships as well as more traditional sources of information; existing information retrieval techniques are adapted to help identify, organize, and track not simply query popularity, but topic popularity; and, information from query logs is utilized to help find preferred queries that express a similar information need. The broader impacts of this work are two-fold. By exploiting the untapped information present in Web search engine query traces, this project increases the understanding of how people search on the Web and for what they are looking. This knowledge is applied to generate algorithms and tools to support searchers as well as those who want to be found by those searchers. These tools, as well as datasets collected or generated, will be made available to the research community via the project Web site (http://www.cse.lehigh.edu/~brian/nsf/queries-03.html).